Many applications acting as Transport Control Protocol (TCP) servers rely upon external means to provide high availability, such as High Availability Cluster Multi-Processing (HACMP™), which is available from International Business Machines. In some HACMP clusters of servers (e.g., in an HACMP cluster of SMTP servers), only one of the servers is active at any moment, while the remaining servers remain on hot standby. This is an inefficient use of hardware resources. Consequently, a need for load balancing arises to optimize the load distribution among such servers. However, whenever there are a plurality of clients, and if the server and the clients are controlled by the same entity (e.g., reside within single data center), there seldom is a compelling reason to deploy dedicated load balancing devices that are known to be expensive.